process supervision and scheduling

[awehttam]

This has been bugging me for a while.

• Right now the documentation provides examples for running the daemons as crontabs. This doesn't provide supervision if the process dies.
• Systemd is becoming defacto but not everyone likes it and it isn't on every distribution. At least crontab is portable.
• On upgrade the daemons should be restarted.
• Scripts like the activity report need to be scheduled. As more reporting and utility scripts (like logrotate) come in to play crontab gets long.

I'm thinking that admin_daemon or binkp_schedule could be used for process supervision and scheduling of common tasks. Then, only that one single daemon needs to be run from crontab or systemd.

The question systemd vs. crontab still comes into play. Crontab I could write a while [ 1 ] wrapper that runs the actual daemon, then only a @reboot entry is needed so that the "supervising daemon" starts on reboot.

Or do we just do systemd at the user level?

I'm trying to avoid ever needing to run a command as root and I like the cron approach because it's easy to manage.

Thoughts?

[awehttam]

I took a stab at writing a supervisory daemon. I know it would work on Linux but I had problems with networking under Windows. It basically killed it for me.

As it stands my feeling is to just leave it up to the sysop for how they want processes to run, and the project can document things as crontabs.

Totally open to feedback on this one

[awehttam]

we're using supervisord in docker